Earticle

현재 위치 Home

Article

Enhancing AI-Driven Dermatology with CycleGAN : Synthetic Clinical-to-Dermoscopic Image Generation

첫 페이지 보기
  • 발행기관
    선문효정학술연구회 바로가기
  • 간행물
    The Journal of Sciences and Innovation for Sustainable Peace(구 The journal of Hyojeong Academia) 바로가기
  • 통권
    Vol. 3 No. 2 (2025.10)바로가기
  • 페이지
    pp.22-28
  • 저자
    Sarreha T. Rikta, Wonsang You
  • 언어
    영어(ENG)
  • URL
    https://www.earticle.net/Article/A481832

※ 기관로그인 시 무료 이용이 가능합니다.

4,000원

원문정보

초록

영어
Dermoscopy and clinical imaging are vital for diagnosing skin conditions, as they pro-vide complementary perspectives that enhance diagnostic insight. However, while clinical images are easy to acquire, dermoscopic imaging often faces limitations due to equipment costs, expertise requirements, and data scarcity. This imbalance restricts the development of robust deep learning models in dermatology. To overcome this challenge, we propose a CycleGAN-based bidirectional image translation framework capable of generating realistic synthetic dermoscopic and clinical im-ages from their respective counterparts. The model effectively preserves key pathological structures while bridging the modality gap between the two imaging domains. Quantitative evaluation demonstrates promising results, with FID scores of 153.93 (clinical) and 117.03 (dermoscopic), and mean LPIPS scores of 0.6368 (clinical) and 0.6421 (dermoscopic), confirming the visual realism and structural consistency of the generated images. By reducing dependence on costly data acquisition and improving dataset diversity, this approach establishes a foundation for integrating synthetic data into dermatological deep learning, ultimately enhancing diagnostic accuracy and clinical ap-plicability.

목차

Abstract
1. Introduction
2. Materials and Methods
2.1. Dataset Preprocessing
2.2. Overview of CycleGAN architecture
2.3. Experimental Setup
3. Results
3.1. Qualitative Evaluation
3.2. Quantitative Evaluation
4. Discussion
5. Conclusion and future work
References

키워드

Dermoscopy Clinical skin image Deep Learning CycleGAN Synthetic image generation

저자

  • Sarreha T. Rikta [ AIIP Lab, Department of Information and Communication Engineering, Sun Moon University, Asan 31460, Korea ]
  • Wonsang You [ AIIP Lab, Department of Information and Communication Engineering, Sun Moon University, Asan 31460, Korea ] Corresponding Author

참고문헌

자료제공 : 네이버학술정보

간행물 정보

발행기관

  • 발행기관명
    선문효정학술연구회 [Sun Moon Hyojeong Academy Society]
  • 설립연도
    2023
  • 분야
    복합학>학제간연구
  • 소개
    Journal of Hyojeong Academia aims to serve as a global platform where researchers and scholars of various disciplines can contribute ideas for our sustainable global community of Co‐existence, Co‐prosperity, and Co‐righteousness. The journal is a multidisciplinary, open‐access, internationally peer‐reviewed academic journal, and it invites all areas of research conducted in the spirit of post materialism including studies centering on God, studies unifying religions and sciences, and studies on all aspects of Co‐existence, Co‐prosperity, and Co‐righteousness.

간행물

  • 간행물명
    The Journal of Sciences and Innovation for Sustainable Peace(구 The journal of Hyojeong Academia)
  • 간기
    반년간
  • pISSN
    2982-9305
  • 수록기간
    2023~2026
  • 십진분류
    KDC 238 DDC 289

이 권호 내 다른 논문 / The Journal of Sciences and Innovation for Sustainable Peace(구 The journal of Hyojeong Academia) Vol. 3 No. 2

    피인용수 : 0건 (자료제공 : 네이버학술정보)

    함께 이용한 논문 이 논문을 다운로드한 분들이 이용한 다른 논문입니다.

      출처 : 네이버학술정보

        0개의 논문이 장바구니에 담겼습니다.

        페이지 저장
        소속기관 조회
        이용자님의 소속기관(단체)이 서비스에 가입되어 있는지 확인해 보십시오.
        기관회원에 소속되어 있는 이용자는 원문을 무료로 이용할 수 있습니다.